Lung Cancer Classification using Convolutional Neural Networks (CNN) has emerged as a critical research endeavor in medical imaging, holding profound implications for early diagnosis and effective treatment. The accurate categorization of lung cancer plays a pivotal role in enhancing patient outcomes and reducing mortality rates. This study presents a comprehensive study leveraging the power of CNNs to achieve robust and high-performing lung cancer classification. The study capitalizes on two distinct datasets, comprising 1097 and 364 lung images, respectively. The methodological progression unfolds with meticulous data scaling, followed by a judicious 80:10:10 data split to facilitate model training, validation, and testing. To address the class imbalance, an innovative approach utilizing Synthetic Minority Over-sampling Technique (SMOTE) is employed, bolstering the learning process on both training and validation sets. The crux of the study lies in the meticulously designed CNN architecture, boasting a stratified composition of 9 layers. Anchored by the quintessential convolutional layers, the model adeptly captures intricate hierarchical features inherent to the input 2D lung images. These acquired representations are seamlessly channeled through dense layers, culminating in the accurate and confident classification of each image into its respective class. The experimental outcomes underscore the potency of the proposed approach, with the first model yielding an impressive accuracy of 99.1%, while the second dataset achieves remarkable perfection with a 100% accuracy rate. This research serves as a significant stride towards the realization of a reliable and efficient tool for lung cancer classification, holding promise for early detection and personalized medical interventions.